I'm a master's student using SPM for the first time, attempting to do EEG source localisation, but despite reading all relevant sections of the SPM12 manual and virtually every lecture or presentation I've been able to find on the internet, I'm still unsure if my approach is correct. Could you please briefly validate it?

I've been really trying to conquer this by myself but admittedly, I'm making no progress any more. I'll be thankful for anything helpful including educational references that are aimed at beginners.

Data: ERP study of rotation-related negativity with two experimental conditions. 61 channels x 200 samples (800 ms) x 28 subjects. I've got everything in a single dataset where individual subjects are represented as trials (which are already averaged trials for every channel) and labeled in D.trials.

Objective: Find the areas of different activation between the two conditions between -400 and -200 ms.

Procedure: Since I'm using the template single_subj_T1.nii MRI and default electrode positions, it is my understanding that I can consider the data from individual subjects as trials from a single subject. Do I need to separate n individual subjects into n datasets, each with two trials (one experimental, one control)?

I open 3D Source Reconstr., select Group Inversion, Model Standard, Time-Frequency contrast yes, Time window (ms) -400 -200, Frequency 0, Power evoked.

SPM does its job and produces two spatial 3D NIfTI files (name_1_t-400_-200_f_[1/2]) along with this graph: http://i.imgur.com/aXmPnZu.png (I assume those are averaged over the -400 to -200 time span, correct?)

It is my understanding that the upper graph shows the estimated power at a given point where it's biggest for the respective condition (colour) while the lower shows all the 512 most prominent dipoles in the brain space, correct? How can I interpret log-evidence? What is considered a reliable result?

Now, I believe those two NIfTI files are contrast images and when I use them as overlays over the MRI template in a viewer such as MRIcroGL, I can clearly see the differences: http://i.imgur.com/kahBLYM.png

Provided that the above approach is correct, could you please suggest further analysis procedure? The SPM12 manual says that it's the same as 2nd level fMRI analysis but a) I don't know how to proceed, and b) none of the provided examples deals with data generated by source localisation but regular fMRI data which appear to be different.

Also, how can I extract the difference between the two contrasts (I assume it should be possible to generate another 3D NIfTI file that would only include the areas of difference between the two input files) and possibly even identify those areas automatically (here I assume there are such utilities)?